Wednesday, May 18, 2016

Poll Roundup And Seat Betting Watch: Neutrally Geared Edition

2PP Aggregate: 50.0-50.0 tie (-0.3 for Coalition since last week)
Coalition would probably just win election "held now" (seat projection 78-68-4*)
* see note re Greens below

So far a quietish week on the polling front, with Morgan and Essential the only national polls expected or seen.  However there are lots of polling snippets to deal with, plus the belated appearance of actual colour in the betting section, so here goes.

The week saw a fairly comfortable audience-scored win to Bill Shorten in a "people's forum" debate watched by hardly anyone, which commentators generally called a draw.   With the audience for debates screened on Pay TV basically containing no-one but political addicts, it's doubtful these debates have much influence on voting intention unless something really out of the box occurs.  The week-to-week shifts in voting intention so far this campaign (and the mock campaign before it) are so small they could be explained by anything ... or nothing.  A campaign about little of consequence meanders along in search of some defining Events.

About the most strategically interesting thing so far is Labor's attempt to tackle negative gearing.  The received wisdom for decades has been that negative gearing ideally should be reformed, but wouldn't because the cost of a policy to reform it would be death by scare campaign.  At this early stage we're seeing no sign of it hurting Labor significantly and perhaps even signs that voters might welcome Oppositions taking a little bit of political risk now and then - even when they're not miles ahead of the government.  We will see if this is still the same in another six weeks' time.

Morgan's poll this week had a really strange (even by Morgan standards) primary result with just 36.5% for the Coalition and 33% for Labor - and supposedly 30.5% for Others, including 15.5% for the Greens and 5% for NXT.  The collective Others polled 21% in 2013, including PUP which has since collapsed.  It's just not possible to take this aspect of the Morgan reading seriously, and it's probably time for another reminder that Morgan is an amalgam of two methods (SMS and face-to-face), both of which nobody else uses.  The last-election 2PP was released as 52-48 for Labor, though those primaries would normally give 53-47.  The respondent-preferences 2PP was 52.5 to Labor.

Essential, which has apparently leant to Labor more or less since Turnbull gave Abbott the shove, produced another 51:49 to ALP.  Essential included the Nick Xenophon Team for the first time, giving them 3%.  After considering all the primaries, I aggregated Morgan at 52.2 to ALP (despite an increasing temptation to just calculate my own 2PPs from Morgan's primaries) and Essential at 50.3 to Coalition (with the point added for apparent skew).  All this took the Coalition down a few tenths of a point and at the time of writing it's a tie.  All square, 50-50.  Here's the smoothed tracking graph:

All up, we've been at more or less 50-50 with no convincing move for five weeks in a row now.  It doesn't really mean that things are all that close, since 50-50 would mean a very likely Coalition win, and the Coalition still holds 78 seats on my seat projection model.

One might think polling would be getting more and more predictive as we get closer to the election, but historically this isn't really so until the last few weeks.  A lot of strange stuff has happened about two months out from elections (most notably the 9/11 attacks in 2001) and polling seven weeks out only explains about 46% of variation in federal election outcomes.  As noted in past roundups I have two regressions which I now and then update - one of which takes into account the Coalition's history of strong performances relative to polling, and the other of which doesn't.

The simple projection now says the Coalition will get 49.9% 2PP (with a standard deviation of 2.1 points) and gives it just a 65% chance of winning, while the more complex one says 50.3% (similar SD) with a 72% chance of a Turnbull return.  It's not a tossup since 50-50 isn't really a tie in terms of winning chances, but three more years for this government is far from a certain thing seven weeks out.  If the projection were a poll, it would have a margin of error of four points, meaning that pretty much anything could still happen and that confidence that this has to be a close election is misplaced.

By the way, seven of the 11 governments polling worse than 50-50 at this stage lost, but all 15 polling better (since 1949) won.

By the way last week there was a Lonergan (50-50) which wasn't aggregated because a pollster has to do three national polls by the same method for me to include it.  However, I'll keep an eye out for more.

The Metapoll aggregate has a new release up with a 50.7% 2PP to the Coalition not counting betting market data, or 50.4% including it.  I am unsure what betting market data could possibly account for the difference.  Metapoll projects a 78-66-6 parliament (2 for the Greens and one for NXT) but is projecting the Green and NXT seats off "seat level betting market data".  I am not sure how seat level data could be used to reach this conclusion since minor party seat odds are notoriously prone to "longshot bias" and are also non-independent of each other.

Other aggregators will be listed here as they update (scores for Coalition): Andrew Catsaras also 50.0, Bludger Track and Luke Mansillo 49.9, Phantom Trend 49.1.

Projecting the Greens

A few weeks back I had a go at finding vote threshholds for the Xenophon Group to win seats.  News that the Coalition may direct preferences to the Greens has fired up the possibility of the Greens winning more than just Melbourne this election.  Batman (where David Feeney apparently had a $2.3 million memory fail that lasted for years - Barry O'Farrell, move over) and Wills (which is a vacancy and therefore a juicy target) are on the frontline of possible Green pickups. Antony Green has an excellent roundup of the seat targets.

I am holding off doing a specific projection for Greens target seats until we see where the preference situation goes, but if the Liberals do decide to preference the Greens, the Greens are really rather likely to pick up a second seat.  Something seen in both the NSW and Victorian state elections was an ability of the Greens to intensify their vote in target seats even if their statewide vote did nothing special.  This makes it very difficult to translate their national vote share to seats.

The possibility of the Liberals preferencing the Greens has created some other commentary.  The detailed Saturday Paper article argues that the aim is to give the Coalition more chance of having the largest number of seats in any hung parliament hence "conferring the first right to form government in the event that neither side wins a clear majority", but Australia doesn't actually have that tradition.  Preferencing the Greens can make sense simply on the basis that Labor then has to waste resources defending seats that would otherwise be easy wins, but the government also wants people thinking that a vote for Labor is a vote for a hung parliament and a rerun of 2010-3.  Every word from the Greens about what they would ask for in such a parliament and who they would govern with assists the Coalition.


Not much more leadership stuff this week, but Essential had Malcolm Turnbull with a net negative rating for the first time, even if his -2 (40-42) was only a net drop of two points since last month.  Bill Shorten is on -9 (34-43, up 5), his best result for a year.  Turnbull leads Shorten as "better Prime Minister" 43-28.

While I've not seen the exact question design or results, there is reporting that when Galaxy asked Western Sydney respondents which leader would do the best for their area, Turnbull was no longer preferred.  It would be useful to see this sort of question repeated nationwide to see if voter preference for incumbent PMs in "beauty contest" polling is driven by an abstract view that they would be good for the nation, rather than for the voter's own area.

Seat Polls

The last week has seen the first serious wave of "neutral" seat polling with Galaxy finding the Coalition ahead in Lindsay (54% 2PP), Herbert (53), Leichhardt (52), Gilmore and Reid (51), level in Dobell and Banks and trailing in Macarthur (49).  Of course most of these results are within the margin of error for a seat poll.  ReachTEL (see separate Tasmanian article as there are some unusual method decisions in these ones) has the Coalition up in Braddon (53% respondent 2PP), Lyons and Bass (51), trailing in Franklin (46) and uncompetitive against Andrew Wilkie in Denison (34, though I don't think the Liberals would be second even on the polled primaries).  The Coalition-held marginals in this polling have an average 2PP swing of 1.8 points to Labor, which is about what should be expected given the current national aggregate, the fact that six are "double sophomore" seats, and the non-inclusion of any WA seats in the list.  Nothing to see here that we shouldn't already know.

My Little Pony (aka other polling)

Essential found 48-32 support for the government's internships policy, with Labor and Greens voters somewhat against but Others voters slightly in favour.  Essential also found that voters tend to think this election campaign is too long.

A coalition of environmental groups released a poll that is another example of the tediously common practice of  left-wing "pony polling" in this country.  Supposedly 64% of voters say they will be more likely to vote for a party if it supports moving to 100% renewable energy within 20 years.  Voters also supposedly say in vast numbers (you can read the figures at the link) that they are more likely to support parties that want more action on climate change, the Barrier Reef and phasing out coal.

Frankly, if this kind of polling worked then the Greens would be in government already, or the major parties would be very much greener, but the majors have been ignoring this kind of polling for ages and have invariably got away with it.  The reason it can be ignored is that the way voters respond to issues when confronted with them cold in isolation is very different to how they react when asked to prioritise issues, or when asked an actually nuanced question that plays the environment off against costs or employment.  Even if responses are representative, a voter who says they are more likely to vote a certain way based on a policy change might be only 0.1% more likely.  (I haven't seen the poll wording, by the way, so I'm assuming the worst about it, but that isn't necessary to the point).  The environment probably won't decide the election, no matter how strong or weak any arguments that it should, and for any party trying too hard for green preferences on the back of a simplistic reading of these sorts of findings, a part of the fate that befell Mark Latham in 2004 awaits.

Seat Betting

The usual disclaimer applies - this section is not about promoting gambling or claiming that markets are predictive, but rather tracking whether they are predictive or not.  It's time, finally, to break out the colour-coded seat betting tracker.

At present these are the seats considered on the Sportsbet market to be changing hands or else reasonably close:

Loss (Coalition to Labor): Barton*, Paterson*
Close Loss (Coalition to Labor): Solomon, Dobell*Eden-Monaro, Macarthur,  Petrie, Capricornia, Hindmarsh, Burt
Tie: Swan
Loss (PUP to Coalition): Fairfax

ALP Close Holds: Batman (vs Grn), Richmond (three-cornered), Lingiari, Adelaide
Coalition Close Holds: Banks, Reid, Page, Gilmore, Lindsay, Robertson, Macquarie, New England (vs IND), Bonner, Brisbane, Forde, Leichhardt, Dickson, Lyons, Bradden, Deakin, Corangamite, Dunkley, La Trobe, Hasluck, Cowan, Stirling.

Odds in Batman (Green at $3) have not yet changed following the Feeney property revelation.

I've made a minor change by including Richmond in the close-seats list on the grounds of the odds of the ALP incumbent being longer than $1.40.  Both the Coalition and Greens are vaguely competitive in this seat.

One change is that Lyons has moved into Coalition favouritism following a ReachTEL that showed a massive Liberal primary lead (with the relatively close 2PP of 51-49 clearly a result of an aberrant respondent-preferences sample).  I have never been convinced that Labor should be favourite for this seat, though at times I have considered it not much better than a tossup for the incumbent, and there is still a realistic chance that it could go.

Whatever happened last week with Lingiari has reversed and there is also movement back to the Coalition in several seats.

Here is a colour-coded tracker of the change in favourites at the times I have checked:

(Red = ALP favourite, blue = Coalition favourite, grey = tie)

The Coalition starts with 91 seats, three of which are notionally Labor, plus Fairfax is a given.  Of their 92 seats including Fairfax, there are five Labor has been favoured to gain at every check, while nine Coalition and two Labor seats have seen changes in favouritism (if anyone has done more detailed tracking they may have seen more).

The Sportsbet 2PP market has finally moved and come down to an average 51.6 to Coalition.  The William Hill exact seats market (in bands of five) is currently pointing to 77 Coalition wins.  There is also an exact seats market in single seats (except for "hung parliament") which has 81 and 82 Coalition seats as joint favourites, but also points to 77 Coalition wins on average.  Possihly both these markets are actually a little more positive for the Coalition than they seem, because of longshot bias on the Labor-win ranges.  It is worth keeping an eye on exact seat markets in single seats as I have found them to be quite predictive in the past.

I am expecting quite a number of polls to be released next week and will post another roundup around this sort of time then.

Friday Updates:

There are two new national polls out on Friday night - a ReachTEL and an Ipsos.  Full details of the ReachTEL are available but for the Ipsos we have just the 2PP: 51-49 to Coalition.

I make the 2PP for the ReachTEL 50.0 by last-election preferences as well so it did nothing to my aggregate, but the Ipsos result puts the Coalition back to 50.1 as an end of week reading and they start the new week at 50.2.

Note that both these results predate the big news story of the last day, the AFP raids on Labor offices and staffers over the NBN leaks.

1 comment:

  1. "Preferencing the Greens can make sense simply on the basis that Labor then has to waste resources defending seats that would otherwise be easy wins, but the government also wants people thinking that a vote for Labor is a vote for a hung parliament and a rerun of 2010-3."

    One other reason for Liberal preferences going to the Greens might be that (especially in the case of Grayndler) Labor losing their safe seats means they may lose future ministers and leadership contenders.